LM Unit Root Test with Panel Data: A Test Robust To Structural Changes

This paper proposes an LM test for the unit root hypothesis using panel data. The LM statistic based on the pooled likelihood function is obtained by standardizing the average of the LM statistic for individual time series. Under the null hypothesis, the statistic follows the standard normal distribution in the limit as N, T goes to infinity as long as N/T approaches any finite number, regardless of whether structural breaks are present. According to the Monte Carlo simulation results, the LM test is robust to the presence of structural breaks, and is more powerful than the popular test proposed by Im, Pesaran and Shin (1997) in the benchmark case where no structural breaks are involved.